diff --git a/source/Canopy-Metrics-Tests/CanopyZincCounterTest.class.st b/source/Canopy-Metrics-Tests/CanopyZincCounterTest.class.st
index 3c6e623..9754eec 100644
--- a/source/Canopy-Metrics-Tests/CanopyZincCounterTest.class.st
+++ b/source/Canopy-Metrics-Tests/CanopyZincCounterTest.class.st
@@ -6,14 +6,30 @@ Class {
 }
 
 { #category : 'as yet unclassified' }
-CanopyZincCounterTest >> testHandleAnnouncementAdjustsDurationWithExponentialMovingAverage [
-	"Mirrors ZincCounter (pharo-metrics, out of scope otherwise): first duration sets the
-	baseline, subsequent durations are folded in with 99% weight on the running average."
+CanopyZincCounterTest >> testAFreshCounterAlreadyCarriesTheUsualStatusCodes [
+	"Parity with what the servers export today: the status series exist from the first scrape at
+	zero instead of appearing with the first response of that kind, and the average duration is a
+	number rather than nil - which is not even valid exposition."
+	| counter |
+	counter := CanopyZincCounter new.
+	self assert: counter duration equals: 0.
+	self assert: (counter responses at: 200) equals: 0.
+	self assert: (counter responses at: 500) equals: 0.
+	self assert: counter responses size equals: 15
+]
+
+{ #category : 'as yet unclassified' }
+CanopyZincCounterTest >> testHandleAnnouncementFoldsDurationsIntoARunningAverage [
+	"Mirrors ZincCounter (pharo-metrics): the average starts at zero and every duration is folded
+	in with 99% weight on the running value. The first request therefore does not set a baseline -
+	it moves the average by a hundredth, which is what the servers have been exporting all along.
+	Canopy did set the baseline from the first request, because its average started as nil; the
+	nil guard in adjustDuration: is unreachable now."
 	[ CanopyZincCounter reset.
-	CanopyZincCounter instance handleAnnouncement: (CanopyZincTestAnnouncement new responseCode: 200 duration: 10; yourself).
+	CanopyZincCounter instance handleAnnouncement: (CanopyZincTestAnnouncement new responseCode: 200 duration: 1000; yourself).
 	self assert: CanopyZincCounter instance duration equals: 10.
-	CanopyZincCounter instance handleAnnouncement: (CanopyZincTestAnnouncement new responseCode: 200 duration: 20; yourself).
-	self assert: CanopyZincCounter instance duration equals: (((99 * 10) + 20) / 100) asInteger ]
+	CanopyZincCounter instance handleAnnouncement: (CanopyZincTestAnnouncement new responseCode: 200 duration: 1000; yourself).
+	self assert: CanopyZincCounter instance duration equals: (((99 * 10) + 1000) / 100) asInteger ]
 		ensure: [ CanopyZincCounter reset ]
 ]
 
diff --git a/source/Canopy-Metrics/CanopyZincCounter.class.st b/source/Canopy-Metrics/CanopyZincCounter.class.st
index 223cc09..c4ff648 100644
--- a/source/Canopy-Metrics/CanopyZincCounter.class.st
+++ b/source/Canopy-Metrics/CanopyZincCounter.class.st
@@ -83,7 +83,19 @@ CanopyZincCounter >> increaseStatusCode: statusCode [
 CanopyZincCounter >> initialize [
 	super initialize.
 	requests := 0.
-	responses := Dictionary new
+	duration := 0.
+	responses := Dictionary new.
+	self initializeResponses
+]
+
+{ #category : 'initialization' }
+CanopyZincCounter >> initializeResponses [
+	"Seed the status codes the servers actually answer, so their series exist from the first
+	scrape instead of appearing out of nowhere with the first response of that kind. Same list
+	as pharo-metrics seeds today - a counter that starts at zero reads differently from one that
+	starts existing halfway through the day."
+	#( 200 201 202 204 301 302 400 401 403 404 405 500 502 503 504 ) do: [ :code |
+		responses at: code put: 0 ]
 ]
 
 { #category : 'installing' }